On Wed, Oct 19, 2016 at 6:44 PM, Heikki Linnakangas <hlinn...@iki.fi> wrote:
> On 10/19/2016 02:32 PM, Heikki Linnakangas wrote: > >> >>> >> Oh, forgot that this needs to be backported, of course. Will do that >> shortly... >> > > Done. > Thanks! > > This didn't include anything to cope with an already-corrupt FSM, BTW. Do > we still want to try something for that? I think it's good enough if we > prevent the FSM corruption from happening, but not sure what the consensus > on that might be.. > > I thought it will be nice to handle already corrupt FSM since our customer found it immediately after a failover and then it was a serious issue. In one case, a system table was affected, thus preventing all DDLs from running. Having said that, I don't have a better idea to handle the problem without causing non-trivial overhead for normal cases (see my original patch). If you've better ideas, it might be worth pursuing. Thanks, Pavan -- Pavan Deolasee http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Training & Services